An executive-level explanation of the modern ranking pipeline. Useful for anyone signing off SEO budget who'd like to understand what they're paying for.
Stage 1 — Crawl
Googlebot discovers URLs through sitemaps, internal links, external links and submitted URLs. It downloads them subject to crawl budget — a per-site allocation based on site authority, server response speed and demand signals.
Stage 2 — Index
Crawled pages enter the indexing pipeline. Content is parsed, deduplicated, classified by topic and stored in a massive distributed index. Not every crawled URL is indexed — quality, duplication and relevance filters apply.
Stage 3 — Retrieval
When a query is issued, the system retrieves candidate pages from the index. This is broad-stroke matching — hundreds or thousands of potential candidates surface for any non-trivial query.
Stage 4 — Ranking
Candidates are scored against ranking signals. The exact signals are proprietary but include: content relevance (semantic, entity-based), authority (link signals), user signals (CTR, engagement), Core Web Vitals, freshness and dozens of others. Different queries weight signals differently — "best laptop 2026" weights freshness heavily; "what is technical seo" weights authority.
Stage 5 — Re-ranking & Personalisation
The top candidates are re-ranked considering user context (location, device, query history) and SERP features (featured snippets, AI Overviews, People Also Ask). The final result you see is personalised — different users see different SERPs.
AI Overviews
Since 2024, AI-generated answers sit above traditional organic results for many queries. They draw from indexed content but compress click-through opportunity. Being cited in AI Overviews is a new ranking objective — and the signals that earn citation overlap with but aren't identical to traditional ranking signals.
The whole pipeline takes milliseconds. The work that earns ranking in it takes months. That mismatch is why SEO compounds slowly and beats most channels long-term.